JP2012198928A - Database system, program, and report retrieval method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a technique which facilitates granting of metadata to an image and input of retrieval conditions, and is high in retrieval accuracy of images and reports.SOLUTION: On the basis of network information in which a plurality of character information in each of classification items among a prescribed number of attribute items are associated in a network shape, a list display is visibly output in which at least one character information is listed for every classification item and a plurality of character information in each of the classification items among a plurality of classification items are associated in a network shape. In the list display, under the condition that each character information is specified for some classification items in a plurality of classification items as a factor which configures a report, the specified character information is regarded as a search keyword group, and the keyword search in which a plurality of character information associated with each image data in a database is regarded as metadata for data retrieval is performed, and consequently one or more report data is detected.

Description

  The present invention relates to information search processing technology.

  In the medical field, a technique for converting a large number of interpretation reports into electronic data and storing them in a database system such as PACS (Picture Archiving and Communication System) has become widespread. In this database system, the interpretation report converted into electronic data and the image data to be interpreted are stored in association with each other.

  By the way, in cases where you obtain reference materials at the time of diagnosis or create data that summarizes representative cases, you can search for similar images based on a certain image or search past interpretation reports. If it is possible, the utility value is very high.

  As a conventional technique for searching for a similar image based on a certain image, a search technique using the feature amount of the image is known, but there is a problem that the search accuracy is low.

  Therefore, a technique for searching for similar images using both the feature amount of the image and the metadata added to the image (for example, Patent Documents 1 and 2), and a technique for easily adding metadata to the image (for example, Patent Documents 3 and 4), and a technique (for example, Patent Document 5) for searching for a moving image based on the similarity of dynamic metadata (dynamic feature amount) have been proposed.

JP 2002-259410 A JP 2005-100090 A JP 2000-298606 A JP 2002-175298 A JP 2001-134589 A

  However, the techniques disclosed in Patent Documents 1 and 2 require a complicated operation for adding metadata to an image. Moreover, in the technique of the above-mentioned Patent Document 3, it is possible to select appropriate metadata from metadata options prepared in advance. The types of possible metadata are limited and the degree of freedom is low. In the technique disclosed in Patent Document 4, it is necessary to input the metadata in advance before registering the metadata in the binary data, and a complicated operation for assigning the metadata is required. Furthermore, in the technique of the above-mentioned Patent Document 5, it is very complicated to input search conditions at the time of search using metadata.

  The present invention has been made in view of the above-described problems, and an object of the present invention is to provide a technique that allows easy addition of metadata to an image and input of search conditions and high search accuracy of images and reports.

  In order to solve the above-mentioned problem, the invention of claim 1 is characterized in that, for each report included in a plurality of reports, a plurality of character information constituting the report and one or more image data corresponding to the report are directly and A plurality of character information is listed for each attribute item included in a related information database in which related information that is indirectly related and / or a predetermined number of attribute items to which the plurality of character information belongs is listed, and An input support information database storing network information in which a plurality of character information items belonging to each classification item are associated with each other between a predetermined number of attribute items, and a part of the network information Based on the network information, at least one character information is included for each of the plurality of classification items included in the predetermined number of classification items. An output means for visually outputting a list display in which a plurality of character information belonging to each of the classification items is associated in a network state and at least one image data among the plurality of classification items; Designating means for designating character information for each of the plurality of classification items in the list display in response to a user operation in a state where the at least one image data is visually output by the output means. Determining means for determining character information designated for each of the plurality of classification items by the designation means as an element constituting a report according to a predetermined report model; and the at least one image by the output means The data is visibly output, and the designation means displays the plurality of segments in the list display. In response to a user action in a state where one or more character information is specified for one or more classification items that are a part of the items, the one or more character information is used as a search keyword group, Visible by the output means by performing a keyword search using the plurality of character information directly and / or indirectly related to each image data in the related information database as metadata for data search. Report detecting means for detecting one or more report data related to the at least one image data output to the image data.

  The invention according to claim 2 is the database system according to claim 1, wherein for each report included in the plurality of reports, a report is associated with one or more image data corresponding to the report. A report database in which information is stored, a character information extraction means for extracting a plurality of character information constituting each report by performing language analysis on each report included in the report database, and a plurality of reports For each included report, a plurality of character information constituting the report extracted by the character information extracting means and one or more image data corresponding to the report are directly and / or indirectly associated and stored. And a storage control means for constructing a related information database.

  The invention according to claim 3 is the database system according to claim 1 or 2, wherein a plurality of character information constituting the related information is added to one or more image data corresponding to each report. The attribute information is included.

  The invention according to claim 4 is the database system according to claim 3, wherein the attribute information includes information indicating patient ID, age, sex, modality, examination site, interpretation doctor, and doctor in charge. It includes at least one piece of information.

  The invention according to claim 5 is the database system according to any one of claims 1 to 4, wherein the related information is described in RDF.

  The invention according to claim 6 is the database system according to any one of claims 1 to 5, wherein each character information included in at least one or more pieces of character information is associated with each character information. A dictionary database in which one or more related character information having the following relationship is stored in association with each other, and the keyword search has the predetermined relationship with respect to one or more character information included in the search keyword group It includes a process of extracting one or more related character information from the dictionary database and adding it to the search keyword group.

  The invention according to claim 7 is the database system according to claim 6, wherein the predetermined relationship includes at least one relationship among synonyms, synonyms, and related terms. .

  The invention according to claim 8 is the database system according to claim 6 or claim 7, wherein the predetermined relationship includes a hierarchical relationship of character information concepts, and the keyword search is performed by the search keyword group. Including one or more superordinate character information extracted from the dictionary database and added to the search keyword group.

  The invention according to claim 9 is the database system according to any one of claims 1 to 8, wherein the predetermined number of classification items includes imaging conditions, parts, basic findings, characteristics, diagnosis, and conclusion. Including at least one item.

  The invention according to claim 10 is the database system according to any one of claims 1 to 9, wherein the keyword search is performed by one of the plurality of classification items in the list display by the specifying means. Among the one or more pieces of character information specified for the above classification items, the search is performed for one or more pieces of character information belonging to a predetermined predetermined classification item included in the predetermined number of classification items. It includes processing that is selectively adopted as a keyword group.

  The invention of claim 11 is the database system according to any one of claims 1 to 10, wherein at least a part of the predetermined number of classification items is classified in response to a user action. Item specifying means for specifying an item is further provided, wherein the keyword search is performed by the specifying means for at least one character information specified for one or more classification items of the plurality of classification items in the list display. Among these, it includes a process of selectively adopting one or more pieces of character information belonging to some of the classification items designated by the item designation means as the search keyword group.

  The invention according to claim 12 is the database system according to any one of claims 1 to 11, wherein the plurality of reports include interpretation reports.

  The invention according to claim 13 is the database system according to any one of claims 1 to 12, wherein data can be transmitted to and received from the server having the report detecting means via the Internet line. A terminal device connected to the terminal device, wherein the specifying means is responsive to a user action on the terminal device, and the output means outputs the at least one image data in a visible manner. In the display, character information is designated for each of the plurality of classification items, and the server transmits one or more report data detected by the report detection means to the terminal device via the Internet line. Having a transmission means, and the terminal device has the output means, wherein the output means uses the transmission means. And outputting one or more reports data transmitted visually.

  The invention according to claim 14 is the database system according to claim 13, wherein the server deletes a predetermined type of personal information from one or more report data detected by the report detection means. The transmission means transmits one or more report data from which the predetermined type of personal information has been deleted by the deletion means to the terminal device via the Internet line.

  The invention of claim 15 is a program for causing the database system to function as the database system according to any one of claims 1 to 14 by being executed by a computer included in the database system.

  The invention of claim 16 is a report search method, wherein (A) for each report included in a plurality of reports, a plurality of character information constituting the report, and one or more image data corresponding to the report, A plurality of character information is listed for each attribute item included in a related information database in which related information directly and / or indirectly related is stored, and a predetermined number of attribute items to which the plurality of character information belongs. And preparing an input support information database in which network information in which a plurality of character information items belonging to each classification item are associated with each other in a network form is stored between the predetermined number of attribute items; B) Classification items for a plurality of classification items included in the predetermined number of classification items based on a part of the network information included in the network information. At least one character information is listed for each, and a list display in which a plurality of character information belonging to each classification item is associated with each other among the plurality of classification items, and at least one image data is output (C) in a state in which the at least one image data is visually output by the output means, in response to a user action, in the list display, A step of designating character information for each category item, and (D) character information designated for each of the plurality of category items in step (C) constitutes a report according to a predetermined report model Determining as an element; and (E) a state in which the at least one image data is visually output by the output means. In response to a user action in a state where one or more character information is specified for one or more classification items that are a part of the plurality of classification items in the list display, A keyword search using the plurality of character information directly and / or indirectly associated with each image data in the related information database as metadata for data search And a step of detecting one or more report data associated with the at least one image data visually output by the output means.

  In addition, “characters” in the present specification include not only ideographic characters such as kanji, but also phonetic characters such as romaji and kana, as well as numerals such as roman numerals.

  In addition, “keyword search” in this specification refers to a search based on a keyword set in advance for each item when searching a database or the like.

  According to the database system of claim 1, character information constituting a report in which detailed information about an image is described is attached to the image as metadata, and the character information attached to the image is used as a keyword. As described above, since the keyword search using the metadata can be performed, a search technique with high report search accuracy and easy input of search conditions can be provided.

  According to the database system of the second aspect, since the character information constituting the report originally associated with the image can be given to the image as metadata, a special addition of metadata to the image is possible. No operation is required. That is, it is possible to easily add metadata to an image.

  According to the database system of the third aspect, it is possible to further improve the search accuracy by including the attribute information originally given to the image in the metadata.

  According to the database system of the fourth aspect, the search using the character information regarding the item characterizing the contents of the image can be performed, so that the search accuracy can be further improved.

  According to the database system of the fifth aspect, by using the RDF, the relationship between the image and the character information constituting the report can be easily described.

  In the database system according to any one of claims 6 to 8, a word search slightly different depending on a report creator is performed by performing a keyword search including a keyword having a predetermined relationship from a keyword given first. The probability that a desired image or report will be detected is increased regardless of the wrinkles.

  According to the database system according to claim 7, by performing keyword search considering synonyms, synonyms, related words, and the like, a desired image or Increases the probability that a report will be detected.

  According to the database system of the eighth aspect, the probability that a desired image or report is detected is increased by performing a keyword search in consideration of a keyword of a superordinate concept.

  According to the database system of the ninth aspect, the character information indicating the characteristic contents related to the image can be given to the image, so that the search accuracy can be improved.

  According to any of the database systems according to claim 10 and claim 11, by performing a search by selectively using character information belonging to some items of character information assigned to an image, the search can be performed. The required load can be reduced.

  According to the database system of claim 11, since it is possible to perform a search by selectively using character information belonging to a desired item among character information assigned to an image, the image is assigned to an image designated by a user. It is possible to perform a search by excluding items of character information that have not been made.

  According to the database system of claim 12, the interpretation report is often databased in association with the image, and the interpretation report includes a lot of characteristic character information of the image, so that it is easy to give metadata to the image. And the search accuracy can be improved.

  According to the database system of claim 13, the user can use the Internet, which can easily increase the amount of information to be searched, for visual information on similar images and related reports for a specified image. Therefore, the probability of obtaining a desired image or report can be further improved.

  According to the database system of the fourteenth aspect, personal information can be prevented from leaking over a wide range.

  According to the program of the fifteenth aspect, effects similar to the first to fourteenth aspects can be obtained.

  According to the image search method of the sixteenth aspect, the same effects as those of the first to fourteenth aspects can be obtained.

It is a figure which shows the outline | summary of the database system which concerns on embodiment of this invention. It is a figure which shows the outline | summary of the information processing content which concerns on an input assistance function. It is a figure which shows the outline | summary of the information processing content which concerns on a search function. It is a figure which illustrates the state which inputs a new interpretation report with a report input device. It is a block diagram which shows the function structure which concerns on the input of a new report, and information retrieval. It is a figure which illustrates single report structured data. It is a figure which shows the example by which the element which concerns on many existing interpretation reports was structured. It is a figure which illustrates the information which linked | related the detailed element by description of RDF. It is a flowchart which shows the operation | movement flow of support information creation operation | movement. It is a figure which illustrates an inspection list screen. It is a figure which illustrates an interpretation report input screen. It is a figure which illustrates the content of the request matter displayed on a request matter display area. It is a figure which illustrates the detailed content of the test | inspection displayed on a test | inspection information display area. It is a figure which illustrates the list of frequent phrases. It is a figure which illustrates the list of summaries. It is a figure which illustrates information acquisition condition determination template display. It is a figure which shows the example of a display of an input assistance template display. It is a figure which shows the example of a display of an input assistance template display. It is a figure which shows the example of a display of an image display screen. It is a figure which illustrates the list | wrist of the command which concerns on a search. It is a figure which shows the content of the character information which shows the characteristic of a search reference image. It is a figure which illustrates the memory content of dictionary DB. It is a figure which illustrates the search result screen of a similar image and a related report. It is a figure which illustrates the setting screen which customizes search conditions. It is a flowchart which shows the creation operation flow of a new report. It is a flowchart which shows the search operation | movement flow of a similar image / related report. It is a figure which illustrates the display screen of a related image. It is a figure for demonstrating the content of a thesaurus. It is a figure which shows the outline | summary of the database system which concerns on a modification. It is a figure which illustrates the input support template display which concerns on a modification. It is a figure which illustrates the dialog for determining the search condition which concerns on a modification. It is a figure which illustrates the search result screen of the image and finding sentence which concern on a modification.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings.

<Outline of database system>
A database system according to an embodiment of the present invention is a system for reporting a result of interpretation of an image such as a radiographic image, and the image is normally stored on an image server of PACS (Picture Archiving and Communication System). There is also a database (report database) that stores reports obtained by interpreting images. This report database (report DB) may exist in a server in the PACS or may exist in another server outside the PACS.

  FIG. 1 is a diagram illustrating an overview of a database system 1 according to an embodiment of the present invention.

  The database system 1 is a system for managing and processing medical information in a hospital, for example, and a server 100 and terminals 10 and 20 are connected to a network line NTW such as a LAN so that data can be transmitted and received between them. . Further, the storage unit 200 is connected to the server 100 so as to be able to transmit and receive data, and further, terminals 30 to 50 are connected to be able to transmit and receive data via the network line LN.

  The storage unit 200 stores a medical information database (medical information DB) 210 that stores medical information (hereinafter also referred to as “medical information”) related to a large number of patients who are medical targets.

  The medical information DB 210 stores an image database (image DB) in which a plurality of image data obtained by radiography in the radiology department for each patient is stored, and each report included in the plurality of interpretation reports is stored in the image DB. A report in which one or more image data and data indicating an interpretation report described for the one or more image data are associated (hereinafter also referred to as “report data with image” or simply “report data”). A database (report DB) and a structured database (structured DB) in which data (single report structured data) in which each report data with images stored in the report DB is structured are stored. Yes.

  Note that image data associated with one report data with an image is not limited to one frame, and may be image data for one frame or more, for example, two frames or more. In the report data with image, data of natural text and various character information constituting the interpretation report is associated with information for specifying the location of the image data (for example, description of file name or URI).

  In addition, each image data stored in the image DB is given information related to an attribute (DICOM attribute) of DICOM (Digital Image and Communications in Medicine). The attribute information includes, for example, information indicating a patient ID, age, sex, modality, examination site, interpretation doctor, and doctor in charge, but includes at least one of the seven pieces of information. It is good also as a structure.

  The medical treatment information DB 210 stores information (examination list information) of a list (examination list) indicating a list of examinations for a large number of patients. In the examination list information, for example, information on DICOM attributes given to the image data for each examination is listed. More specifically, the examination list information includes, for example, patient attributes (patient ID, patient name, date of birth, age, gender) related to each examination, a state indicating whether an interpretation report has been created, and examinations. Information indicating the examination ID to be identified, the examination date, the part corresponding to the examination object, the modality indicating the imaging condition, and the number of images taken is included. The medical care information DB 210 stores attribute information indicating ordering (request items) from the doctor in charge and details of the examination (information on the patient and examination) in association with the examination ID.

  The server 100 writes / reads various information to / from the medical information DB 210 and performs information processing based on the various information stored in the medical information DB 210. Information processing in the server 100 is appropriately managed by a data management function installed in the terminals 10-50.

  In the database system 1, for example, the terminals 10 and 20 are terminals used by the doctor in charge of the patient, the terminal 30 is a terminal (report input device) 30 from which a radiology interpreting doctor inputs an interpretation report, and the terminals 40 and 50 are radiation. This terminal is used by a department engineer.

  When an order is received from the terminals 10 and 20 to the radiology terminals 40 and 50, a radiology engineer acquires image data of the affected area using an MR apparatus, a CR apparatus, or the like. The image data of the affected part is stored in the medical care information DB 210 from the terminals 40 and 50 via the server 100. Thereafter, a radiology doctor inputs an interpretation report in the report input device 30 while referring to a screen on which image data stored in the medical information DB 210 is visually output.

  The server 100 uses the interpretation report (existing interpretation report) already stored in the medical treatment information DB 210 to create information (input support information) that supports the input of a new interpretation report, and is necessary from the input support information. Information is acquired (extracted) and provided to the report input device 30 in the form of a template display.

  Further, the server 100 constructs the structured DB in the medical treatment information DB 210 by structuring using the report data with images stored in the report DB. Then, when a new interpretation report is input in the report input device 30, the server 100 uses the structured DB and uses image data similar to the image data specified in the report input device 30 or a related interpretation report (for example, The data of a finding sentence constituting a part of the interpretation report is retrieved as reference information and provided to the report input device 30. The structured DB plays the role of information (search support information) that supports search.

  A function for supporting the input of a new interpretation report in the server 100 is hereinafter referred to as an “input support function”. The input support function includes a function (search function) for searching for image (similar image) data similar to image data designated by the user and related interpretation report (related report) data. The input support function including the search function will be described later.

  2 and 3 are diagrams showing an overview of information processing contents related to the input support function in the database system 1, and FIG. 3 is a diagram focusing on the information processing contents related to the search function.

  The medical information DB 210 stores data indicating captured images and created interpretation reports. As shown in FIG. 2, the report input device 30 is a device that acquires image data to be interpreted from the medical treatment information DB 210 and creates data indicating an interpretation report. Then, the server 100 reads out the data indicating the existing interpretation report from the medical treatment information DB 210 and analyzes it by the analysis function, thereby creating information (input support information) for supporting the creation and input of the interpretation report. Stored in the DB 111. In the report input device 30, a new interpretation report is input using the input support information. Further, the newly created data indicating the interpretation report is added to the medical care information DB 210 and used to reinforce the input support information stored in the input support DB 111.

  As shown in FIG. 3, the medical treatment information DB 210 stores a report DB 211 in which a large number of report data with images are accumulated, an image DB 212 in which a large number of image data capturing an affected area is accumulated, and a report DB 211. A structured DB 213 in which single report structured data in which each image-attached report data is structured is stored. The report input device 30 can obtain image data similar to the designated image data or data related to the interpretation report from the medical information DB 210 as a search result. Note that data indicating a new interpretation report input from the report input device 30 is also used to reinforce the report DB 211 and the structured DB 213.

  FIG. 4 is a diagram illustrating a state in which a new interpretation report is input in the report input device 30. As shown in FIG. 4, in the report input device 30, a radiology interpreting doctor appropriately refers to a screen (image display screen) 401 that visually outputs patient image data stored in the medical information DB 210. A new interpretation report is input on the input screen 402. FIG. 4 shows a state in which image data is visually output to the left half display area LH of the image display screen 401 and no image data is visually output to the right half display area RH. The display area RH functions as an area for displaying various search results in the search function.

<Input support function>
The input support function uses a large number of report data with images stored in the medical information DB 210 as past knowledge, extracts elements necessary for inputting a new interpretation report from the many report data with images, and performs RDF (Resource It has a function of creating input support information and presenting it appropriately by structuring using Description Framework). The input support function also has a function of performing a search using search support information (here, the structured DB 213).

  The operations realized by the input support function are mainly operations for creating input and search support information (support information creation operation), operations related to selection of a test corresponding to a new interpretation report to be input (test selection operation) , And an operation that actually supports the input of a new interpretation report based on the input support information (input support operation) and an operation realized by the search function (search operation).

<Functional configuration>
FIG. 5 is a block diagram showing a functional configuration relating to input of a new interpretation report using the input support function including the search function and the input support function in the database system 1. Among the functional configurations shown in FIG. 5, the functions of the control units 120 and 31 are the programs stored in the storage units 110 and 34 each configured by a hard disk or the like in the server 100 and the terminal (report input device) 30. Functions executed by the control units 120 and 31 configured by a CPU or the like are conceptually shown. Note that various types of data temporarily stored in various types of information processing executed by the functional configuration are temporarily stored in, for example, RAMs built in the control units 120 and 31 as appropriate.

  The server 100 mainly includes a storage unit 110 and a control unit 120.

  The storage unit 110 stores various programs and various data for realizing various controls and functions in the server 100. For example, the storage unit 110 stores an input support DB 111 and a dictionary database (dictionary DB) 130.

  The input support DB 111 stores input support information created by a support information creation operation. Specific contents of the input support information will be described later.

  The dictionary DB 130 is a database in which each character information such as a word or phrase is associated with a word having the same meaning as the character information, a similar meaning or a related meaning, and the like. That is, in the dictionary DB 130, character information such as words and phrases having a predetermined relationship such as synonyms, synonyms, and related words are associated with each other. In other words, the dictionary DB 130 is obtained by associating each piece of character information included in at least one piece of character information with one or more pieces of character information (related character information) having a predetermined relationship. Note that the dictionary DB 130 may be associated with words having one or more relationships among synonyms, synonyms, and related words.

  The control unit 120 performs overall control of various controls and functions of the server 100, and includes a data reading unit 121, an inspected determination unit 122, a data construction unit 123, an information search unit 125, a report configuration unit 127, and a data document. The embedding unit 128 is provided as a function.

  The data reading unit 121 reads report data with image, attribute information corresponding to image data related to the report data with image, and examination list information from the medical treatment information DB 210. The data reading unit 121 sends the report data with image, the attribute information, and the inspection list information to the inspected determination unit 122 in the support information generation operation and the input support operation, and in the inspection selection operation, the inspection list Information is sent to the control unit 31 of the report input device 30.

  The inspection completion determination unit 122 refers to the inspection list information to determine whether or not an interpretation report has been created by writing a finding sentence or the like for the interpretation report of each inspection. Then, in the support information creation operation, the inspected determination unit 122 sends the generated report data with image (that is, inspected) together with the attribute information to the data construction unit 123, and in the input support operation, incomplete (that is, incomplete) Unreported image-attached report data is sent to the control unit 31 of the report input device 30 together with the attribute information.

  The data construction unit 123 extracts a necessary element from a finding sentence that is a natural sentence that constitutes an interpretation report represented by the report data with an image for a plurality of report data with an image, and includes various kinds of information included in the attribute information. Elements are also extracted and structured using RDF, so that input support information is created and the input support DB 111 is constructed, and search support information is created and the structured DB 213 is constructed.

  The input support information created here is that each element belonging to the item of the plurality of elements is mutually networked between the items included in the items of the plurality of elements (here, character information) constituting the finding sentence. It is information (hereinafter also referred to as “element network information”) including associated information (hereinafter also referred to as “network information”).

  The search support information created here includes, for each report included in a plurality of reports, a plurality of character information (for example, words / phrases) constituting a finding sentence of the report and one or more image data corresponding to an interpretation target of the report Is a database (structured DB 213) constructed by accumulating information (hereinafter also referred to as “related information”). The structured DB 213 is also referred to as a related information database (DB).

  In response to a request from the report input device 30, the information search unit 125 uses a part of the element network information stored in the input support DB 111 to configure a template display that directly supports the input of the interpretation report. Network information (hereinafter also referred to as “partial network information”) is acquired. Then, the information search unit 125, based on the partial network information, data related to a display screen including a list display of partial network information in which each element belonging to each of the plurality of element items is associated with each other (network Display screen data) is generated and output to the control unit 31.

  Further, the information search unit 125 searches for and acquires similar image data and related report data related to the image data specified in the report input device 30 using the structured DB 213 as a search target, and displays the search result. Data (search result display screen data) is generated and output to the control unit 31. In this search operation, one or more character information associated with one or more designated image data is used as a search keyword, and a plurality of character information associated with the image data in the structured DB 213 is assigned to the image data. It is regarded as metadata for data search, and a search process (that is, a keyword search process) is performed according to whether or not the metadata matches the search keyword.

  The report composition unit 127 receives new single report structured data input from the control unit 31 in the form of data structured in accordance with the operation of the operation unit 32 by the user in the input support operation, Based on the report structured data, new report data with images in accordance with a predetermined rule is generated.

  The data writing unit 128 adds the report data with image generated by the report configuration unit 127 to the report DB 211 in the medical treatment information DB 210. At this time, the new single report structured data input to the report composition unit 127 is added to the structured DB 213 in the medical care information DB 210 in the same format. Further, the new single report structured data input to the report composition unit 127 is sent to the data construction unit 123 as it is, and is also used for updating the input support DB 111.

  The report input device 30 mainly includes a control unit 31, an operation unit 32, a display unit 33, and a storage unit 34.

  The control unit 31 performs overall control of various controls and functions of the report input device 30, and includes a task management unit 311, a display control unit 312, and a query generation unit 313 as functions.

  When an examination (report input target examination) that is a target for inputting a new interpretation report is selected, the task management unit 311 identifies a task related to creation of new report data with an image corresponding to the report input target examination. .

  The display control unit 312 causes the display unit 33 to visually output various screens based on various information input from the server 100. For example, the network display screen is visually output on the display unit 33 based on the network display screen data input from the information search unit 125, or the search result display screen is visible on the display unit 33 based on the search result display screen data. Output.

  The query generation unit 313 generates a query using the character information associated with the specified image data as the user's interpretation doctor operates the operation unit 32 in various ways, and sends the query to the information search unit 125. Output.

  The operation unit 32 includes a keyboard, a mouse, and the like, and outputs various signals to the control unit 31 by being appropriately operated by a user.

  The display unit 33 is configured by various displays such as a liquid crystal display. For example, the display unit 33 visually outputs network display image data and search result display screen data generated by the information search unit 125 according to control by the display control unit 312. .

  The storage unit 34 stores various programs and various data for realizing various controls and functions in the report input device 30.

  Hereinafter, the four main operations realized by the input support function, that is, the support information creation operation, the examination selection operation, the input support operation, and the search operation will be described in order with reference to FIG.

<Support information creation operation>
First, the data reading unit 121 reads report data with an image, attribute information corresponding to the report data with the image, and examination list information from the medical treatment information DB 210, and sends them to the inspected determination unit 122. The inspection completion determination unit 122 refers to the inspection list information to determine whether or not report data with an image has been created for each inspection, and the generated report data with an image is sent to the data construction unit 123 together with attribute information. Send it out. Then, the data construction unit 123 extracts necessary elements from the finding sentence, which is a natural sentence of the interpretation report, and also extracts various elements included in the attribute information, and performs structuring using RDF. The structuring of the report data with an image is realized by using machine learning, for example.

  Hereinafter, machine learning and structuring of report data with images in the data construction unit 123 will be described.

  First, for example, a learning corpus or the like is given as teacher data to learn information serving as a structuring reference.

  The learning corpus includes a large amount of text data according to the finding sentence format (sentence model) of the interpretation report. This sentence model shows, for example, the structure of the interpretation sentence of the interpretation report, such as imaging conditions → parts → basic findings (feature-conclusion) → diagnosis (diagnosis-conclusion). In the learning corpus, for example, the classification item name of each element (here, each character information) constituting the sentence sentence model included in the interpretation report model (also referred to as “report model”) is tagged to each word. It is attached.

  For example, the classification items of elements include “diagnosis content category (hereinafter abbreviated as“ category ”)”, “imaging condition”, “part”, “basic findings”, “diagnosis content (hereinafter abbreviated as“ diagnosis ”), and the like. .

  In the learning corpus, for example, the element classification item name “category” is tagged to phrases such as “cerebral infarction”, “ischemic change”. The element classification item name “imaging condition” is tagged to phrases such as “T1 image”, “T2 image”. The element classification item name “part” is tagged to phrases such as “frontal lobe”, “temporal lobe”. The element classification item name “basic findings” is tagged to phrases such as “spot-like, patchy high-signal areas”, “dot-like high-signal areas”, and so on. The element classification item name “diagnosis” is tagged for phrases such as “old cerebral infarction”, “broad old cerebral infarction”, etc.

  In addition, when such a learning corpus is given from the outside to the data construction unit 123, data indicating a typical sentence model in which a conclusion is specified to some extent in order to create new report data with images, which will be described later. Is also given from the outside.

  The machine learning function of the data construction unit 123 extracts a word from the learning corpus and stores it for each classification item of the corresponding element. That is, in the data construction unit 123, teacher data including a learning corpus is used as a learning material, and words belonging to each element classification item are learned and stored for each element classification item with reference to the teacher data. Here, the data that is learned and stored is used as model data (model data) that indicates what element classification items should be decomposed into each element constituting the existing interpretation report.

  Further, the identification function of the data construction unit 123 is based on the model data learned as described above, and the classification item of the element and the actual phrase used for the interpretation report input to the data construction unit 123 And identify.

  If the machine learning method described above is used, element classification items can be identified only for elements (here, words) given in advance in teacher data, but the following machine learning method is used. As a result, it is possible to identify the element classification items even for elements not previously given in the teacher data.

  For example, the machine learning function of the data construction unit 123 decomposes the learning corpus into morphemes by morpheme analysis, and for each morpheme, the morpheme itself, the morpheme part of speech, the morpheme utilization form, the front and rear (for example, two before and after) A pattern in which a morpheme belonging to a certain classification item appears is learned using information or the like. Then, the identification function of the data construction unit 123 can recognize the classification item of the element even for an element (in this case, character information such as a phrase) that has not been given in advance according to the pattern.

  More specifically, for example, in the machine learning function of the data construction unit 123, sentences in the form of “along **” (various words are entered in the “**” part) frequently appear in the learning corpus. In this case, when a word indicating a part frequently appears in the “**” part, it is possible to learn a pattern in which a word indicating the part is included in the “**” part. If the identification function of the data construction unit 123 uses such a pattern, the phrase “pituitary gland / along” from the phrase “pituitary gland / along” is determined from the context of the word for an existing interpretation report. Can be extracted. The machine learning can be realized by using a so-called SVM (Support Vector Machine). Such machine learning improves the accuracy of natural language processing.

  In this way, the data construction unit 123 performs language analysis on each report data with image included in the report DB 211, thereby extracting a plurality of character information constituting each report data with image.

  Then, in the data structuring function included in the data construction unit 123, based on the information identified by the identification function, various kinds of information included in the report data with image and the attribute information are classified for each element classification item (element). The information related to the report data with image is structured by decomposing it into RDF and describing it in RDF.

  FIG. 6 is a diagram illustrating data (single report structured data) in which elements of one interpretation report (including elements constituting attribute information related to an existing interpretation report) are structured. As shown in FIG. 6, element classification items “category”, “imaging condition”, “part”, “basic findings”, and “diagnosis” are assigned to the attribute data “A-0001” of the report data with image by RDF. The attribute values “cerebral infarction”, “T2WI”, “frontal lobe”, “high signal area”, and “lacuna infarction” are described in association with each attribute. The attribute value “high signal area” is described with the element classification item “feature” as the attribute item and the attribute value “light” associated with the attribute.

  Further, the attribute items “examination part”, “modality”, “age”, and “gender” relating to the attribute information of the attached data and the image data are attributed to the file “A-0001” of the report data with image by RDF. The image data file names “img15.jpg”, “head (SKULL)”, “MR”, “53”, and “male (M)” are described in association with each other as attribute values. Note that the attribute items (attribute items) to be associated in the single report structured data are not limited to the eleven classification items shown in FIG. 6, for example, other attribute items included in the examination attribute information. Or a predetermined number of classification items including other items constituting a natural sentence such as “Conclusion”. In FIG. 6, only one image data is associated, but a plurality of image data may be associated.

  Also, here, the attribute values of the attribute items related to the attribute items other than the image data and the attached image are directly associated with the file “A-0001” of the report data with image. The value is indirectly associated with the file name “A-0001”. However, the present invention is not limited to this, and the image data is directly associated with the attribute values of some or all of the attribute items. May be. For example, the attribute values “head (SKULL)” “MR” “53” “male (M)” for the attribute items “examination part”, “modality”, “age”, and “sex” related to the attribute information of the image data are: The attribute value “img15.jpg” indicating the file name of the image data may be directly associated.

  In addition, the data construction unit 123 converts the single report structured data as shown in FIG. 6 into a number of examined images stored in the medical information DB 210 by the machine learning function, the identification function, and the data structuring function. Create each report data.

  In this way, for each piece of report data with an image included in a plurality of report data with an image stored in the report DB 211, a plurality of character information constituting the report data with an image is associated with an interpretation target of the report data with an image. By storing a large number of single report structured data created by associating with one or more pieces of image data obtained in the medical information DB 210, related information DB corresponding to search support information for supporting search operation, that is, structured DB 213 Is built.

  In addition, the data construction unit 123 arranges a plurality of attribute values associated in each single report structured data for each attribute item, reassociates them with each other, and describes them in RDF. Do for all of the data. As a result of this processing, a plurality of attribute values (here, character information such as phrases) are enumerated for each attribute item included in the plurality of attribute items, and the attribute values are mutually exchanged between the attribute items. Associated information (in network form) (network information) is created. By storing the network information in the storage unit 110, information (input support information) that supports input of a new interpretation report, that is, the input support DB 111 is constructed.

  Here, the structured DB 213 and the input support DB 111 can be easily described by using RDF.

  FIG. 7 is a diagram illustrating an example in which elements related to a large number of existing report data with images are structured, and illustrates network information about the examination site “head (SKUL)” and “chest (CHEST)”. ing. In FIG. 7, the associated words are connected by a solid line. In FIG. 7, in order to avoid complication of the figure, the phrase is appropriately described as “XX”, etc., and the solid line indicating the association is also illustrated at a relatively upper part in the figure, The solid line is omitted.

  Further, when creating the network information in the data construction unit 123, the number of word combinations between the items associated in each single report structured data is counted, and the count information is stored in the input support DB 111. Store it. For example, the number of combinations of phrases such as “head-MR-53-M-cerebral infarction-T2WI-frontal lobe-high signal area-lacuna infarction" is stored in the count information.

  In the network information stored in the input support DB 111, what words are described in the existing interpretation report is shown in a form in which the words are associated with each other between the attribute items. Therefore, it is effective to output the network information visually and use it when inputting a new interpretation report. In particular, it is effective to provide each attribute item as an item of an element (input element item) and as a candidate (option) for inputting a plurality of words listed in each attribute item in the form of a template display.

  By the way, in the network information stored in the input support DB 111, if there are many synonyms (for example, “T2 weighted image” and “T2WI”, etc.) for the words listed in each attribute item, the number of options increases too much. Designation becomes difficult. Therefore, for the synonym, when the phrase is detected by the identification function of the data construction unit 123, a process of replacing it with one representative phrase is performed. In addition, when there are “conclusions” etc. in the attribute item, if there are many fluctuations in the expression of the concluding remarks (for example, “think” and “think”), the number of choices will increase and it will be difficult to specify the choices. . Therefore, processing for replacing the expression shake with one representative expression when the conclusion is detected by the identification function of the data construction unit 123 is performed.

  Such replacement of a word with a typical word can be realized by including a table in which a plurality of words and typical words are associated with each other in the teacher data. Regarding the shake of the expression, the shake of the expression may be normalized and replaced with the most frequently used one. Also, all of the synonyms may be replaced with the most frequently used ones.

  Moreover, in the interpretation report, for example, for the attribute item “basic finding”, a phrase (basic finding phrase) in which a major classification and a modifier indicating a feature are combined is often used. For example, the basic finding phrase “light high signal area” is a combination of a large classification “high signal area” and a feature “light”.

  In this way, the basic finding phrases (for example, “light high signal area”, “dense high signal area”, etc.) are composed of a large number of combinations of large classifications and features, so that the number of options increases too much, and the options are not specified. It becomes difficult. Therefore, for example, for the attribute item “basic findings”, the elements constituting the basic finding phrase are decomposed into two-element items “major classification” and “feature”.

  The process of decomposing this component into two-element items is realized by using a well-known SVM or the like on the basis of a model that lists language elements belonging to each element item. The model used here can be obtained by machine learning using teacher data including a corpus in which a large number of basic finding phrases are listed, for example. In this corpus, item names are tagged to a large number of language elements constituting a basic finding phrase.

  Then, for the attribute item “basic findings” in the network information, the basic finding phrase related to a certain broad classification term is replaced with the major classification term. At this time, apart from the network information, as shown in FIG. 8, detailed elements such as characteristics (ie, modifiers) combined to form a basic finding phrase for a large classification term are described by RDF description. Configure the associated information (detailed information). The detailed information forms structured data (detailed information structured data) in which detailed elements (detailed elements) are associated with detailed elements and detailed elements (detailed elements).

  Specifically, in FIG. 8, the detailed elements “none”, “light”, and “dark” belonging to the item “feature” of the detailed element are associated with the term “high signal range” of the large classification that is the element with detailed information. It has been.

  Note that, for example, a phrase belonging to a certain item such as a basic finding phrase also corresponds to a detailed element for a broadly classified term corresponding to an element belonging to the certain item. Therefore, the “detailed element” described below includes detailed elements that constitute a phrase such as features, but also includes a phrase that is a result of combining a plurality of detailed elements as appropriate.

  Further, the network information, the count information, and the detailed information are information indicating associations between elements (herein, phrases) belonging to a plurality of items among a plurality of items, and hence are appropriately referred to as “element network information”. The count information is information (also referred to as “related strength information”) indicating the strength of association between a plurality of elements including one or more detailed elements.

  FIG. 9 is a flowchart showing an operation flow of the support information creation operation. This operation flow is realized by executing a program for realizing the input support function in the control unit 120 of the server 100.

  In step ST1, report data with an image is read together with attribute information from the medical treatment information DB 210 by the data reading unit 121 and the examination completion determination unit 122.

  In step ST2, a finding sentence that is a natural sentence indicated by the report data with an image is structured by the function of the data construction unit 123.

  In step ST3, the data construction unit 123 adds attribute information to the data structured in step ST2, and creates single report structured data as shown in FIG.

  In step ST4, the single report structured data created in step ST3 is stored in the medical information DB 210. The structured DB 213 is constructed by repeating the storage process of the single report structured data in step ST4.

  In step ST5, the data construction unit 123 determines whether single report structured data has been created for all the report data with images stored in the report DB 211. Here, if single report structured data has not been created for all report data with images, the process returns to step ST1, and the next report data with images is read together with attribute information, and single report structured data is obtained. Created and stored. If single report structured data is created for all report data with images, the process proceeds to step ST6.

  In step ST6, the data construction unit 123 creates network information as shown in FIG. 7 based on the single report structured data for all report data with images.

  In step ST7, the network information created in step ST6 is stored in the storage unit 110, whereby the input support DB 111 is constructed, and this operation flow ends. At this time, element network information including network information and count information is created and stored in the input support DB 111.

  By the way, the report data with image stored in the medical treatment information DB 210 increases every time new report data with image is generated according to the input from the report input device 30. It is effective if the interpretation reports accumulated over time can be used as past knowledge. In particular, when an unprecedented finding sentence or the like is included in a newly accumulated interpretation report, it is more effective because the past knowledge further evolves. Although this point will be described later, in the input support operation (described later), past knowledge has evolved in response to the fact that the element to be actually input is specified for each item constituting the observation text of the interpretation report To do.

<Inspection selection operation>
Next, an operation (inspection selection operation) for displaying a screen (examination list screen) showing an examination list and selecting an examination for report input in the report input device 30 will be described.

  The examination selection operation is realized by an input support function, and the data reading unit 121 reads examination list information from the medical information DB 210 and transfers it to the control unit 31 of the report input device 30. The control unit 31 causes the display unit 33 to display the examination list screen based on the examination list information by the function of the display control unit 312.

  FIG. 10 is a diagram illustrating an examination list screen G1. As shown in FIG. 10, the examination list screen G1 includes information on examinations (specifically, patient ID, patient name, date of birth, age, sex, state, examination ID, examination date, examination part, modality, This is a screen displaying a list of (number of images).

  In this examination list screen G1, it can be seen that the examination with the state “uninterpreted” is an examination for which no interpretation report has been created.

  On the examination list screen G1, a thick frame cursor CS1 surrounding one examination is displayed. This thick frame cursor CS1 is moved up and down in response to the operation of the operation unit 32 by the user, and when a determination button (for example, a return key) of the operation unit 32 is pressed in a state where it is adjusted to a desired inspection, One examination surrounded by the thick frame cursor CS1 is selected as the report input target examination. In this way, when a report input target examination is selected, the screen displayed on the display unit 33 transitions from the examination list screen G1 to a report input screen G2 (described later).

  In this examination selection operation, for example, when a report input target examination is selected on the examination list screen as shown in FIG. 10, an element (for example, SKULL) belonging to the item “examination part” and an element ( For example, a combination with CR) is simultaneously designated and recognized.

  When the report input target examination is selected, the task management unit 311 identifies a task related to the creation of a new interpretation report corresponding to the report input target examination. At this time, under the control of the task management unit 311, the request items corresponding to the report input target examination and the attribute information indicating the details of the examination are read from the medical information DB 210 by the data reading unit 121, and the control unit of the report input device 30 31 is provided.

<Input support operation>
Next, the case where the report input device 30 displays a screen for inputting a new interpretation report (interpretation report input screen) and inputs a new interpretation report will be described. Data relating to various screens displayed on the display unit 33 of the report input device 30 is input from the server 100, and the display control unit 312 in the report input device 30 displays the screen on the display unit 33 based on the input data. Realize.

  FIG. 11 is a diagram illustrating an interpretation report input screen G2. As shown in FIG. 11, the interpretation report input screen G2 mainly includes a request item display area A1, an examination information display area A2, a memo area A3, a private memo area A4, a creation support information display area A5, and an observation sentence display area. A6, a frequent phrase display area A7, a summary display area A8, an image attachment area A9, and a command display area A10.

  The requested item display area A1 is an area in which the contents of requested items related to the report input target inspection are displayed. Specifically, based on the information related to the requested items acquired from the medical care information DB 210 via the server 100, for example, the contents of the requested items as shown in FIG. 12 are displayed in the requested item display area A1.

  The examination information display area A2 is an area where the detailed contents of the examination related to the report input object examination are displayed. Specifically, based on the information related to the details of the examination acquired from the medical care information DB 210 via the server 100, for example, the detailed contents of the examination as shown in FIG. 13 are displayed in the examination information display area A2.

  The memo area A3 is an area in which free opinions and other information are described in accordance with the operation of the operation unit 32 by the image interpretation doctor.

  The private memo area A4 is an area in which free opinions and other information are described according to the operation of the operation unit 32 by the radiogram interpreter, similarly to the memo area A3. However, the information described in the private memo area A4 can be browsed only by a specific device or a specific person. For example, browsing can be restricted by filtering using an IP address, input of a password, or the like, which can be used for information management of highly confidential personal information.

  The creation support information display area A5 is an area for visually providing a part of network information acquired (extracted) from the network information stored in the input support DB 111 described above in order to support creation of an interpretation report. Although details will be described later, in the creation support information display area A5, first, a screen for designating an element belonging to the item “category” is displayed, and network information corresponding to the element belonging to the item “category” designated on the screen is displayed. Is obtained from the input support DB 111 and visually output in the form of a template display (hereinafter also referred to as “input support template screen”). The input support template screen will be described in detail later.

  The observation sentence display area A6 is an area in which the observation sentence constituting the interpretation report is displayed, and the observation sentence is written in response to the operation of the operation unit 32 by the interpretation doctor.

  The frequent phrase display area A7 is an area where a list of fixed phrases (frequent phrases) frequently used in finding sentences is displayed. In the frequent phrase display area A7, for example, a list of frequent phrases as shown in FIG. Is displayed. The interpreting doctor appropriately operates the mouse or the like of the operation unit 32 to designate one frequent phrase from the frequent phrase list and copy (fill in) the frequent phrase designated at the cursor position in the finding sentence display area A6. be able to.

  The summary display area A8 is an area in which a list of fixed phrases (summary phrases) frequently used in the finding sentence is displayed. In the summary display area A8, for example, a list of fixed phrases as shown in FIG. 15 is displayed. Is done. The interpreting doctor appropriately operates the mouse of the operation unit 32 to specify one fixed sentence from the list of fixed sentences and copy (fill in) the fixed sentence specified at the cursor position in the finding sentence display area A6. be able to. Specifically, for example, when the standard sentence “There are no special findings” in the summary display area A8 is specified, the character string “No special findings” is copied to the observation sentence display area A6. The

  Information on frequent phrases and fixed phrases may be stored in advance in the storage unit 110 or the like. For example, one or more frequent phrases and fixed phrases may be combined with elements belonging to the two items “examination part” and “modality”. A list of frequent phrases and fixed phrases corresponding to combinations of elements belonging to the two items “examination part” and “modality” specified on the examination list screen are stored in correspondence with each other, and a frequent phrase display area A7 and a summary display area If it is displayed at A8, the convenience is high.

  The image attachment area A9 is an area to which a characteristic and representative image is attached, and a desired representative image is attached in accordance with the operation of the operation unit 32 by an interpreting doctor. For example, among the images displayed in the left half area LH of the image display screen 401 shown in FIG. 4, place the mouse pointer on the desired image and hold down the left mouse button while moving the mouse pointer to the image attachment area A9. A desired image can be attached to the image attachment area A9 by moving (dragging) and releasing the left button of the mouse.

  The command display area A10 is an area in which icons for inputting commands (here, buttons BT1 and BT2) are listed. Specifically, the button BT1 confirms the findings and the contents of the attached image that make up the interpretation report created on the interpretation report input screen G2, registers new image-attached report data in the medical information DB 210, and performs the next interpretation. This is a button for inputting a command to shift to creation of a report (that is, report data with an image). The button BT2 is a button for inputting a command for clearing the findings and the contents of the attached image constituting the image interpretation report created on the image interpretation report input screen G2 and recreating the content of the image interpretation report.

  Next, an operation for inputting a finding sentence constituting an interpretation report using a template display displayed in the creation support information display area A5 will be described.

  In the creation support information display area A5, the network information created as described above in the support information creation operation is visually output in the form of a template display, and options for elements in the template display (here, character information such as phrases) are selected. Is appropriately specified, and a finding sentence is input (that is, a new interpretation report is input).

  However, if a plurality of elements listed in each attribute item are selected as options, and the entire network information (entire network information) is simply provided in the form of a template display as shown in FIG. 7, the number of options listed. Is too large to make choices. Therefore, in the database system 1 according to the embodiment of the present invention, for example, the interpretation doctor who is a user determines elements for some attribute items in the entire network information, so that a part of the network information (one Part network information) is acquired (extracted) and displayed. As described above, in the template display in which a part of network information is displayed as a list, the number of options listed is limited to some extent, so that the options are easy to see.

  Further, for example, for the broad classification term “high signal range” belonging to the basic finding item, terms indicating various characteristics are added to form various basic finding phrases. For this reason, many basic finding phrases relating to one major classification term belong to the basic finding classification item. However, when all of the many basic finding phrases (for example, “light high signal area”, “dark high signal area”, etc.) are displayed in a list, the number of options to be listed is large, and selection of options is difficult.

  For such a problem, it is conceivable to newly provide a classification item related to a detailed element such as “feature” in the list display, but in such a measure, the number of items in the list display increases too much, It will be hard to see. Furthermore, regarding the classification item related to the detailed element, since it is not an essential item to input the detailed element, it is difficult to understand what is the minimum essential item for creating the report.

  Therefore, in the list display in the database system 1, first, for a plurality of basic finding phrases related to the same major classification term, the major classification term is shown as a representative, and in response to a predetermined instruction, the major classification term is changed. Detailed information regarding the plurality of basic finding phrases is displayed. With such a configuration, it is possible to easily input detailed information while clarifying the skeleton of the sentence of the finding sentence (minimum required item).

  Here, a method for acquiring a part of network information from the entire network information will be described.

  FIG. 16 shows a template display (information acquisition condition determination template display) TP1 for determining conditions (hereinafter also referred to as “information acquisition conditions” and “extraction conditions”) for acquiring desired partial network information from the entire network information. It is a figure illustrated.

  As described above, when an examination is selected by the examination selection operation, on the interpretation report input screen G2 (for example, FIG. 11), for example, the contents of the requested items (for example, FIG. 12) are displayed in the requested information display area A1. Detailed contents (for example, FIG. 13) of the examination related to the report input target examination are displayed in the display area A2. Then, by the function of the information search unit 125, the information acquisition condition determination template display TP1 is displayed in the creation support information display area A5.

  In the information acquisition condition determination template display TP1, words (options) are listed for some attribute items “category” of all attribute items in the entire network information. The choices related to the attribute item “category” enumerated here are, for example, elements and items “modality” belonging to the item “examination part” designated at the same time as the selection of the examination for report input in the examination selection operation in the entire network information. ”Is an element related to the item“ category ”associated with the combination with the element belonging to“ ”.

  For example, when the combination of the element “CHEST” belonging to the item “examination part” and the element “CR” belonging to the item “modality” is designated in the examination selection operation, the information acquisition condition determination template display TP1 shows that in FIG. As shown, the elements “lung”, “soft part”, “mediastinum”, “pleura (margin)”, “bone”, “other” belonging to the item “category” are listed as options.

  Information on options related to the item “category” listed in the information acquisition condition determination template display TP1 may be stored in advance in the storage unit 110 or the like. For example, one or more elements related to the item “category” may be stored in two items “ The item “category” corresponding to the combination of the elements belonging to the two items “examination part” and “modality” specified in the examination list screen is stored in correspondence with the combination of the elements belonging to the examination part and “modality”. It suffices that the list of choices related to is displayed on the information acquisition condition determination template display TP1.

  In this information acquisition condition determination template display TP1, when the image interpretation doctor performs various operations on the operation unit 32 to perform a predetermined operation (for example, double click) with the mouse pointer MP set to a desired option, the item “category” is displayed. It is possible to specify options belonging to.

  Here, a combination of options determined on the examination list screen (for example, FIG. 10) and the information acquisition condition determination template display TP1 is designated as the information acquisition condition. For example, in the information acquisition condition determination template display TP1 shown in FIG. 16, when the option “pleura” belonging to the item “category” is designated, the element “CHEST” of the item “examination site” and the element “CR” of the item “modality” ”And the item“ category ”element“ pleura ”are designated as information acquisition conditions. This information acquisition condition is given from the report input device 30 to the information search unit 125 by the control unit 31.

  In response to the specification of the information acquisition condition from the report input device 30, the information search unit 125 searches the input support DB 111 according to a predetermined information acquisition rule. Information for displaying a list of partial network information corresponding to a predetermined information acquisition rule and information acquisition condition as a template display in the creation support information display area A5 from the entire network information (hereinafter also referred to as “display network information”). To obtain). Specifically, from the entire network information, partial network information narrowed down to items specified by a predetermined information acquisition rule is acquired from the network information satisfying the information acquisition condition.

  At this time, the information search unit 125 can extract partial network information satisfying the information acquisition condition from the entire network information by referring to the count information stored in the input support DB 111. At this time, a part of the count information corresponding to the extracted part of the network information is also extracted from the count information.

  The partial network information and the partial count information are collectively referred to as “partial element network information” as appropriate, and the partial network information and the partial count information are appropriately referred to as “network information” and “count information”, respectively. Abbreviated. Here, if the partial network information is extracted in a format described in RDF, the partial element network information and network information after acquisition can be easily described.

  Then, a template display in which the partial network information extracted as described above is visually displayed as a list is provided. Note that the template display provided here supports the input of a finding sentence of a new interpretation report, and is hereinafter also referred to as “input support template display”.

  FIG. 17 is a diagram illustrating a display example of the input support template display TP2. In FIG. 17, the element “MR” for the item “modality”, the element “SKULL” for the item “examination site”, and the element “for the category“ category ” The input support template display TP2 displayed when the combination of “infarct” is designated as the information acquisition condition is illustrated. Various inputs and designations in an input support template display TP2, which will be described later, are performed in accordance with operation inputs of the operation unit 32 by an interpretation doctor.

  As shown in FIG. 17, in the input support template display TP2, the information acquisition condition SC is described at the top, and items “parts”, “basic findings”, and “diagnosis” are sequentially arranged from the left in the area centered at the approximate center. A plurality of words F1 to F3 are displayed. Specifically, for the attribute item “part”, a plurality of words (“frontal lobe>”... “Bridge”) F1, and for the attribute item “basic findings”, a plurality of words (“dotted, patchy”). "High signal area" ... "scattered high signal area") F2, and for the attribute item "diagnosis", a plurality of phrases ("old cerebral infarction" ... "multiple cerebral infarction") F3 Enumerated.

  For the three items “parts”, “basic findings”, and “diagnosis”, the words associated with each other are connected to each other based on partial network information extracted according to the information acquisition conditions (here, , Broken lines). In this way, some network information that has been narrowed down to a certain extent is visually displayed as a list.

  Further, in FIG. 17, a phrase with “>” such as the phrase “high signal range>” indicates a broadly classified term representing a plurality of basic finding phrases and part phrases. In the input support template display TP2, by placing the mouse pointer MP on such a phrase with “>” and right-clicking, a pop-up list format list appears, and detailed elements are designated as appropriate. When a desired phrase is selected, the selected desired phrase is applied to blank fields (blank fields W1, W2, etc.) corresponding to the phrase. Note that frequently used phrases are displayed in the pop-up list that appears when the mouse pointer MP is placed on a word with “>” and the right button of the mouse is pressed, and the mouse pointer MP is moved to the display. Thus, a phrase that is frequently used may be easily selected. Further, when a desired phrase is selected by appropriately specifying detailed elements, as shown in FIG. 18, in the input support template display TP2, the broadly classified terms are selected as desired phrases (here, The configuration may be replaced with “light high signal range>”).

  When a word corresponding to each item is designated, the sentence model display MD of the finding sentence of the interpretation report displayed at the bottom of the input support template display TP2 is referred to. This text model display MD is displayed based on report model data (here, text model) given from the outside together with teacher data and the like.

  This sentence model display MD shows the sentence model of the finding sentence of the interpretation report such as “part-basic findings-conclusion. Diagnosis—conclusion”. Specifically, a sentence model such as “[blank column W2] [pull-down list PD1] [blank column W3] [pull-down list PD2] in [blank column W1]” is shown. In the blank columns W1 to W3, the phrases related to the elements “parts”, “basic findings”, and “diagnosis” are applied, and in the pull-down lists PD1 and PD2, the conclusions related to the basic findings and the conclusions related to the diagnosis are respectively applied. .

  That is, according to the phrase designation by the interpretation doctor, one or more desired phrases (for example, “right frontal gray matter”) among the plurality of phrases (options) F1 are applied to the blank column W1, and the plurality of phrases (options) F2 One of the phrases (for example, “high signal area”) is allocated to the blank column W2, and one of the plurality of phrases (options) F3 (for example, “acute cerebral infarction”) is allocated to the blank column W3. Will be. The order in which words are specified in particular for these three items does not matter.

  Further, in the sentence model display MD, for a conclusion constituting a natural sentence along a predetermined sentence model other than the three items, a desired option can be designated by appropriately matching the mouse pointer MP with the pull-down lists PD1 and PD2. it can. FIG. 18 shows a state in which the word “acknowledge” is designated for the basic findings, and the word “suspect” is designated for the diagnosis. It should be noted that there may be three or more options related to the conclusion, and it may be configured so that other words and phrases constituting a natural sentence such as an adjective and an adverb can be selectively specified.

  As described above, with the words (including phrases as appropriate) designated for the three items and the two conclusions, the interpretation doctor performs various input operations on the operation unit 32, and the mouse pointer MP is set to “add text”. When a predetermined operation (for example, left click) is performed in accordance with the button BT11 to instruct, the control unit 31 determines words designated for each item and the like as elements constituting a report according to a predetermined report model. Then, the finding sentence of the interpretation report displayed on the sentence model display MD is copied to the finding sentence display area A6.

  On the other hand, when the mouse pointer MP is set to the button BT12 for instructing “clear text” and a predetermined operation (for example, left click) is performed, all elements displayed on the text model display MD are cleared.

  A new finding is entered in the finding sentence display area A6 using the input support template display TP2 or the like shown in FIG. 18, and the mouse pointer MP is aligned with the button BT1 at the bottom of the interpretation report input screen G2 to perform a predetermined operation. When (for example, left click) is performed, a new interpretation report registration instruction is issued. Then, the report component 127 is controlled by the elements constituting the new interpretation report including the finding sentence displayed in the finding sentence display area A6 and the information indicating the location of the image data attached to the image attachment area A9. Input from the unit 31. At this time, as for the elements (report constituent elements) constituting the findings of the interpretation report displayed on the sentence model display MD, each report constituent element is designated for each item. Are input in a format such as single report structured data shown in FIG.

  The report construction unit 127 generates new report data with an image according to a predetermined rule based on the new single report structured data and the like, and adds it to the report DB 211 via the data writing unit 128. At this time, new single report structured data is added to the structured DB 213. On the other hand, the data construction unit 123 updates the input support DB 111 by adding new single report structured data information to the input support DB 111.

<Search operation>
As described above, in the input support template display TP2, it is possible to easily input a finding sentence. However, at the time of inputting a finding sentence, elements can be easily selected for some items constituting the finding sentence. However, in a situation where the knowledge is ambiguous and the interpretation is unclear, the remaining items are Selection is difficult. In such a case, if it is possible to refer to an image similar to the image to be interpreted included in the past case and the finding sentence described about the image, it will be a reference for judgment, and input will be easier. Become.

  Hereinafter, a search operation for searching for and presenting a similar image to an image to be interpreted and a finding sentence described for the similar image will be described.

  FIG. 19 is a diagram illustrating a display example of the image display screen 401 displayed on the report input device 30.

  As shown in FIG. 19, when inputting a finding sentence, image data to be interpreted is visually output in the left half area LH of the image display screen 401. FIG. 19 shows a state in which images related to 12 pieces of image data are displayed.

  When the image interpretation doctor operates the mouse included in the operation unit 32 in various ways, when the mouse pointer MP is set to an appropriate position in a plurality of images displayed in the region LH and the right button of the mouse is clicked, the result is shown in FIG. A list of commands CL1 appears. FIG. 20 shows a list of commands CL1 displayed for the image focusing on the image on which the mouse pointer MP is placed.

  The command list CL1 includes a plurality of commands including four commands of “search for similar images”, “search for similar images by selecting attributes to be used”, “display detailed attributes of images”, and “display history”. It is. Then, the image interpretation doctor operates the mouse included in the operation unit 32 in various ways, and the desired command can be designated by clicking the left button of the mouse with the mouse pointer MP set to the desired command.

  First, a case where the command “similar image search” is designated among a plurality of commands in the list CL1 will be described.

  In the input support template display TP2, some of the classification items among a plurality of (here, five) classification items “parts”, “basic findings”, “conclusion”, “diagnosis”, and “conclusion” related to elements constituting the finding sentence. If “similar image search” is specified in a state where an element (here, character information) is specified, one or more images displayed in the region LH are used as search criteria (hereinafter “search”). Also designated as “reference image”. That is, search reference image data (search reference image data) is designated.

  When the search reference image is specified, the character information indicating the characteristics of the search reference image that the character information already specified for some of the classification items is associated with the search reference image in the input support template display TP2. (Also referred to as “image characterization character information”). Further, the image characterization character information appropriately includes character information constituting the DICOM attribute of the search reference image and the attribute value of the information acquisition condition SC.

  FIG. 21 is a diagram illustrating image characterization character information.

  In FIG. 21, character information is not specified for the attribute item “diagnosis”, and the attribute items Pf1 to Pf8 (here, the character information “age”, “sex”, “modality”, “examination part”, “imaging condition”, “part” ”“ Basic findings ”“ features ”) attribute values Pv1 to Pv8 (here, character information“ 53 ”“ male ”“ MR ”“ SKULL ”“ T2WI ”“ frontal lobe ”“ high signal range ”“ light ”) Are associated with each other.

  When the search reference image is specified as described above, the control unit 31 sends the information search unit 125 a similar image related to the search reference image using a plurality of character information constituting the image characterization character information as a search keyword group. A query is sent requesting a search. The term “search keyword group” is used in a concept including not only two or more pieces of character information but also one piece of character information.

  In response to acceptance of the query from the control unit 31, the information search unit 125 performs image characterization characters for a predetermined classification item by keyword search using text information for the structured DB 213 in the medical information DB 210 as a search target. One or more single report structured data having an attribute value similar to information is detected.

  The “predetermined classification item” referred to here is a predetermined predetermined classification item among a predetermined number of classification items to be associated in the single report structured data. Examples of the predetermined partial classification items include the classification items Pf3 to Pf7 shown in FIG. 21, that is, the five classification items “modality”, “examination region”, “imaging condition”, “region”, and “basic findings”.

  As described above, in the keyword search in the information search unit 125, among a plurality of pieces of character information associated with one or more search reference image data (search reference image), a predetermined number to be associated in the single report structured data. One or more pieces of character information belonging to a predetermined predetermined category item among the category items are selectively adopted as a search keyword group. As described above, by selectively using character information belonging to some items of the character information assigned to the image data, the load required for the keyword search process is reduced.

  Further, in the keyword search in the information search unit 125, in order to prevent omission of search, the dictionary DB 130 is referred to and one or more character information included in the search keyword group and a predetermined relationship (synonyms, synonyms, related words, etc.) A keyword search is performed after extracting one or more pieces of character information having a relationship) (hereinafter also referred to as “related character information”) and adding them to the search keyword group. FIG. 22 is a diagram exemplifying the storage contents of the dictionary DB. A combination of “old cerebral infarction” and “old cerebral infarction” having a synonym relationship, and “T2WI” and “T2WI emphasized image”. A combination with a “T2WI image” is illustrated.

  Since the single report structured data includes the information indicating the location of the image data and the file name of the report data with the image, the information search unit 125 detects one or more single report structured data. Then, one or more pieces of image data and one or more pieces of report data with images are uniquely detected. At this time, since the image data detected by the information search unit 125 has an attribute value similar to the attribute value given to the search reference image data, a different image from the search reference image data and a similar image to the search reference image data is displayed. Is detected as image data. Further, since the report data with image detected by the information search unit 125 indicates an interpretation report described for a similar image with respect to the search reference image, the interpretation report data related to the report data with image is included in the search reference image data. It is detected as data of related reports (ie, related reports).

  Then, a search result display screen is provided on which the similar image and the related report data detected as described above are visually output.

  FIG. 23 is a diagram illustrating the display contents of the search result display screen. The search result display screen is displayed in the display area RH on the right half of the image display screen 401.

  In the search result display screen shown in FIG. 23, combinations of similar images, attribute information, and finding sentences related to two report data with images are displayed in order from the top. Specifically, the similar image related to the first report data with image is set as the similar image 1, and two images attached to the report data with image, attribute information (48 years old, female, SKULL, MR) And a finding sentence (in the T2WI-weighted image, high signal areas are observed in the frontal and occipital lobes. Old cerebral infarction is suspected). Below the similar image 1, the similar image related to the second image-attached report data is set as the similar image 2, and the two images attached to the image-added report data, attribute information (68 years old, male, SKULL, MR) and observation text (in T2WI, a faint high signal area is recognized in the frontal lobe and right temporal lobe. Please check carefully by CT).

  The interpreting doctor refers to the search result display screen as shown in FIG. 23, and compares the similar image with the image to be interpreted, and specifies the character information with reference to the past finding sentence. Appropriate character information can be easily specified for the classification item “diagnosis” which has been difficult.

  Next, a case where the command “select an attribute to be used and search for similar images” among a plurality of commands in the list CL1 is specified will be described.

  When the command “Search for similar images by selecting an attribute to be used” is specified in the list CL1, a setting screen for customizing search conditions as shown in FIG. 24 is displayed in the right half display area RH of the image display screen 401. Is displayed.

  In the setting screen shown in FIG. 24, a predetermined number (here, 9) of classification items “age”, “gender”, “modality”, “examination region”, “imaging condition”, “ “Parts”, “basic findings”, “features”, and “diagnosis” are listed, and check boxes are added to the left of each classification item. In this setting screen, the interpretation doctor who is the user operates the mouse of the operation unit 32 in various ways, aligns the mouse pointer MP with a desired check box, and clicks the left button of the mouse, thereby classifying items used for the search. While the check box attached to the left of the box is in the black square state, the check box attached to the left of the classification item not used for the search is in the white square state, and the decision button is moved with the mouse pointer MP. By pressing, a keyword search with customized search conditions can be executed.

  As for the age, if only similar images and related reports related to patients of the same age as that of the patient related to the search reference image are searched, the information obtained as a search result is very little or very biased. It may become a thing. For this reason, in the setting screen shown in FIG. 24, a certain range can be set for the age.

  For example, in the setting screen shown in FIG. 24, there are states in which five classification items “modality”, “examination part”, “imaging condition”, “part”, and “basic findings” are designated as the classification items used for the search. It is shown.

  In addition, since no keyword search can be executed when there is no classification item to be used for the search, a part of the classification items including at least one of the predetermined number of classification items is specified on the setting screen. Otherwise, the keyword search is not executed even if the enter button is pressed.

  In the keyword search in the information search unit 125, among a plurality of pieces of character information associated with one or more search reference image data, setting is made among a predetermined number of classification items to be associated in the single report structured data. One or more pieces of character information belonging to some classification items set on the screen (FIG. 24) are selectively adopted as a search keyword group. In this way, by performing a keyword search by selectively using character information belonging to a desired classification item among character information assigned to search reference image data, for example, an interpreting doctor still has character information. The search can be performed by appropriately excluding items of character information that have not been input.

  FIG. 25 is a flowchart showing an operation flow for creating a new report. Here, when a predetermined operation is performed in the report input device 30, first, an examination selection operation is started, and the process proceeds to step ST11 in FIG.

  In step ST11, the examination list screen (for example, the examination list screen G1 in FIG. 10) is displayed on the display unit 33.

  In step ST12, it is determined whether or not a test (report input target test) for creating a new interpretation report has been selected on the test list screen. Here, the determination in step ST12 is repeated until the report input target inspection is selected, and when the report input target inspection is selected, the process proceeds to step ST13.

  In step ST13, an interpretation report input screen G2 is displayed. At this time, the creation support information display area A5 of the interpretation report input screen G2 includes an item “category” corresponding to the combination of the element of the item “examination part” and the element of the item “modality” relating to the already selected report input target examination. A screen (for example, information acquisition condition determination template display TP1 in FIG. 16) on which elements (here, character information) belonging to “” are listed is displayed.

  In step ST14, it is determined whether or not the character information related to the item “category” is selected in the information acquisition condition determination template display TP1. Here, the determination in step ST14 is repeated until the character information related to the item “category” is selected, and when the character information related to the item “category” is selected, the process proceeds to step ST15. At this time, the information acquisition condition is determined by selecting the character information related to the item “category”.

  In step ST15, in response to the input of the information acquisition condition, partial network information is acquired from the entire network information stored in the input support DB 111. At this time, some count information is also acquired.

  In step ST16, the input support template display TP2 in which the partial network information acquired in step ST15 is visually output is displayed on the display unit 33.

  In step ST17, an element is designated for each item in the input support template display TP2, and other information is input. Here, in a state in which character information (including phrases as appropriate) is designated for each item and each conclusion by an operation input of the operation unit 32 by the interpretation doctor, the mouse pointer MP is set to the button BT11 for instructing “add sentence”. When a predetermined operation is performed, the finding sentence of the interpretation report displayed on the sentence model display MD is copied to the finding sentence display area A6.

  In step ST18, it is determined whether or not a new interpretation report registration instruction has been issued. Here, the processes of steps ST17 and ST18 are repeated until a new interpretation report registration instruction is given, and the mouse pointer MP is put on the button BT1 at the bottom of the interpretation report input screen G2 to perform a predetermined operation (for example, left click). Is performed, it progresses to step ST19.

  In step ST19, a new interpretation report is constructed in the report construction unit 127, and new report data with images is recorded in the medical treatment information DB 210 by the data writing unit 128. At this time, the new single report structured data input in the input support template display TP2 is added to the structured DB 123. The new single report structured data is also used as data for reinforcing the information stored in the input support DB 111 by the functions of the report composition unit 127 and the data construction unit 123. That is, every time new single report structured data is input, the count information is updated while the data is added to the entire network information stored in the input support DB 111.

  FIG. 26 is a flowchart showing a search operation flow for similar images and related reports. This operation flow starts when the process proceeds to step ST16 in FIG. 25, and proceeds to step ST31 in FIG.

  In step ST31, the control unit 31 determines whether or not a search reference image data is specified by operating the operation unit 32 by the image interpretation doctor and an instruction (search instruction) for executing a keyword search is issued. For example, in the command list CL1 as shown in FIG. 20, “similar image search” is designated, or in the list CL1, “select similar attribute and select similar image” is designated in FIG. When the search condition is customized on the setting screen as shown in FIG. 6, a search instruction is issued. Here, the process of step ST31 is repeated until a search instruction is issued, and when a search instruction is issued, the process proceeds to step ST32.

  In step ST <b> 32, a query requesting a search for similar images and related reports with respect to the search reference image is transferred from the control unit 31 to the information search unit 125. When the search condition is customized, the customized search condition information is also transferred to the information search unit 125.

  In step ST33, the information search unit 125 recognizes an attribute, that is, a classification item, that is used for a keyword search that is set in advance or customized.

  In step ST34, the information search unit 125 recognizes the search condition. Specifically, character information belonging to the classification item recognized in step ST32 among the attribute values (here, character information) associated with the search reference image is recognized as a search keyword.

  In step ST35, the information search unit 125 detects similar images and related report data that match the search conditions recognized in step ST34 by keyword search.

  In step ST 36, the search result display screen data related to the search result is generated by the information search unit 125 based on the similar image and the related report data detected in step ST 35, and transferred to the control unit 31.

  In step ST37, the search result display screen is displayed on the display unit 33 based on the search result display screen data generated in step ST36 under the control of the display control unit 312. That is, one or more image data and one or more report data detected in step ST35 are visually output.

  As described above, in the database system 1 according to the embodiment of the present invention, character information constituting a report (herein, an interpretation report) in which detailed information about image data is described is metadata for image data. As given. Then, for example, the image interpretation doctor who is the user simply designates the image to be interpreted as an image to be used as a search reference (search reference image) during interpretation, and the character information already assigned to the search reference image can be obtained. As a keyword, a similar image and related report data related to the search reference image data are detected by keyword search using the metadata. In this way, the text search information describing the detailed characteristics of the image is used as metadata, and the search condition is determined simply by specifying the image, thereby improving the image search accuracy. The search condition can be easily input.

  In particular, interpretation reports are often associated with images and stored in a database, and interpretation reports contain a large amount of character information that represents the detailed characteristics of images, so that metadata can be added to image data and search accuracy can be improved. It can be easily achieved.

  In addition, since the character information constituting the report data with an image originally associated with the image data is automatically added to the image data as metadata, a special operation for adding metadata to the image Is no longer necessary. Therefore, it is possible to easily add metadata to an image.

  Furthermore, the character information given as metadata to the image data is character information indicating characteristic contents regarding the image such as imaging conditions, parts, basic findings, features, diagnosis, and conclusions. Therefore, the search accuracy can be improved.

  Further, by including attribute information related to DICOM attributes or the like originally assigned to the image data in the metadata, the search accuracy can be further improved. For example, by including in the metadata attribute information relating to items that characterize the content of the image (for example, patient ID, age, sex, modality, examination site, doctor in charge, etc.), character information relating to the item that characterizes the content of the image is used. The search accuracy can be further improved.

  In addition, using the dictionary DB 130, a keyword search including keywords having a predetermined relationship (for example, a relationship of synonyms, synonyms, related terms, etc.) is performed from a keyword given first. For this reason, the probability that a desired similar image or related report data is detected is increased regardless of the wording slightly different depending on the creator of the interpretation report.

<Modification>
As mentioned above, although embodiment of this invention was described, this invention is not limited to the thing of the content demonstrated above.

  For example, in the above embodiment, both the similar image and the related report are detected. However, the present invention is not limited to this, and at least one of the similar image and the related report may be detected and presented to the user. .

  In the above embodiment, when creating a new image interpretation report, a characteristic representative image among the plurality of images displayed in the display area LH is displayed in the image attachment area of the input support template display TP2. Although described as being attached to A9, the present invention is not limited to this. For example, in the case where a characteristic representative image is not attached to the image attachment area A9, the image is composed of a plurality of images displayed in the display area LH. The report data with an image in which the entire image group is associated with the interpretation report is generated. In such a case, the single report structured data shown in FIG. 6 may be associated with, for example, a folder name indicating the location of the entire image group. Then, when report data with an image associated with the entire image group is detected by keyword search, all images constituting the image group are displayed as a group of similar images on the search result display screen.

  In the above embodiment, the similar image and related report data search for the image data to be interpreted is performed, but the present invention is not limited to this. For example, data retrieval of similar images and related reports related to image data indicating the progress of a patient who is an interpretation target may be performed. Examples of the configuration include the following configurations.

  For example, when the command “display history” is designated in the list CL1 shown in FIG. 20, a screen showing information of report data with images indicating the past progress of the patient who is the interpretation target as shown in FIG. (History display screen) is displayed in the display area RH. In FIG. 27, the information related to the previous report data with image, the attribute information, and the information including the finding text are shown as the related image 1. Also, an image related to the previous report data with an image, attribute information, and information including a finding sentence are shown as the related image 2.

  Then, in the history display screen shown in FIG. 27, as in FIG. 20, the command list appears by placing the mouse pointer on the desired image and clicking the right button of the mouse, and the command “ Specify "Search for similar images" or "Search for similar images by selecting attributes to use", and similar images and related images related to images displayed on the history display screen by the same operations and operations as in the above embodiment. A report is detected and presented on the report input device 30. In such a configuration, one or more related images that have already been inspected are designated as search reference images and are directly and / or indirectly associated with the image data of the search reference images in the structured DB 213. Character information is designated as a search keyword group.

  For example, when almost no interpretation report is input for an image currently being interpreted, there is little character information associated with the image data to be interpreted. Therefore, when a similar image or related report data relating to image data to be interpreted is directly searched, the character information included in the search keyword group is too small, and the keyword search accuracy is reduced. Therefore, a configuration that detects similar images and related report data related to image data indicating the past progress for which an interpretation report has already been created, that is, similar image and related report data related to image data to be interpreted is indirectly By adopting a configuration that automatically searches, it is possible to improve the accuracy of keyword search. Note that the above configuration improves the accuracy of the keyword search even when there is no significant change between the image showing the past progress and the image currently being interpreted. It is effective against.

  In the above-described embodiment, similar images and related report data related to image data to be interpreted are directly or indirectly searched. However, the present invention is not limited to this. For example, an image that has already been inspected (that is, an interpretation report has been created) is designated as a search reference image, and a large number of single report structured data included in the structured DB 213 is used as a search target. A configuration having a search function in which report data is detected and presented in the report input device 30 may be employed. In such a configuration, when an already inspected image is designated as a search reference image, one or more associated directly and / or indirectly with the image data of the search reference image in the structured DB 213. Is designated as a search keyword group.

  By adopting such a configuration, it is possible to collect a lot of images similar to the images related to each case, and select and collect representative images showing each case. It can be easily created.

  Note that the teaching file is intended to make it easy for doctors to refer to typical cases and characteristic cases, and once created, it is extremely useful for doctors' education and diagnosis decisions. Although effective, it is usually very time consuming and expensive to create. In addition, once created, various maintenance is required as needed.

  Here, not only is it easy to create a teaching file, but it is also possible to easily enrich the description content of the teaching file by collecting a lot of reports related to the images related to each case. The above configuration is also effective during maintenance such as updating the contents of a teaching file.

  The above configuration can also be used effectively in the following training mode.

  For example, when image data serving as a model is given at the time of medical training, and character information indicating the content of the image data is associated, the image serving as a model is used by using the search function as described above. Other image data with the same symptoms as the data can be pulled. Therefore, after conducting a diagnosis while viewing the image related to the model image data, the side receiving the training searches the similar image and the related report related to the model image data and searches the user himself / herself. The results of diagnosis and general diagnosis can be compared. Thereafter, image and report data indicating how the medical condition changes can also be searched.

  In the above embodiment, the dictionary DB 130 stores words and phrases having relationships such as synonyms, synonyms, and related words, but is not limited to this. For example, Phrases may be stored in association with each other, phrases or phrases having similar meanings may be stored in association with each other, or phrases or phrases having related meanings may be stored in association with each other. That is, in the keyword search, a so-called ontology that follows various relationships may be used.

  Furthermore, the dictionary DB 130 may include a database in which the hierarchical relationships of character information concepts such as a thesaurus in which various terms are arranged hierarchically are organized. Hereinafter, a search operation when the thesaurus 19 in which words are arranged hierarchically and the dictionary DB 130 includes the thesaurus 19 will be described.

  FIG. 28 is a diagram illustrating a hierarchical structure of the thesaurus 19. As illustrated in FIG. 28, the highest hierarchy in the hierarchical structure of the thesaurus 19 is a major classification, and words included in the first hierarchy, the second hierarchy, and the third hierarchy are sequentially from each of the major classifications. Branched. In the thesaurus 19, words are systematized based on the hierarchical relationship of concepts, conceptually higher words (higher words) are positioned higher in the hierarchical structure, and conceptually lower words (lower words) are hierarchically structured. The words are arranged hierarchically so that they are located at the lower level. For example, when a certain large classification represents “disease”, “digestive disease” is displayed in the lower layer, “stomach disease” in the second layer, “stomach disease” in the second layer, “ Words related to “gastric ulcer” are arranged. The position of the thesaurus 19 on the hierarchical structure (hereinafter abbreviated as “hierarchical position”) is specified by a thesaurus code which is a symbol given according to a predetermined rule. For example, the hierarchical positions of the words related to the above-mentioned “disease”, “digestive disease”, “stomach disease” and “gastric ulcer” are “A”, “A-10”, “A-10-20” and “A”, respectively. -10-20-30 "is specified by the thesaurus code. Here, “A” is an index that identifies the major classification, and numbers such as “10”, “20”, and “30” are indices that indicate the hierarchical position within the same hierarchy. The hyphen “-” is an index indicating the transition between layers. In the thesaurus 19, since a specific word is allowed to have a plurality of broader terms, the hierarchical position may be specified by a plurality of thesaurus codes.

  By using such a thesaurus 19, for one or more character information included in the initial search keyword group, for example, character information belonging to a layer higher by one or more layers, that is, one or more higher-order character information is thesaurus. A configuration may be adopted in which a keyword search using the search keyword group is performed after being extracted from 19 and added to the search keyword group. In this way, if a keyword search is performed in consideration of keywords of a higher concept, the probability that a desired image or report is detected is further increased.

  In the above embodiment, an example of supporting the input of an interpretation report has been described. However, the present invention is not limited to this, and the present invention is not limited to this, as long as it supports the input of a report described about an image. It can be applied to any report. For example, the present invention can be applied to other medical information reports such as incident reports and diagnostic reports, and to reports in general with images such as sales reports and complaint countermeasure reports.

  In the above embodiment, the predetermined number of classification items constituting the single report structured data includes all of the six classification items “imaging condition”, “part”, “basic findings”, “feature”, “diagnosis”, and “conclusion”. Although it was included, it is not restricted to this, The structure which employ | adopted the 1 or more classification item suitably among the said 6 classification items may be sufficient.

  In the above embodiment, various inputs are made by various operations of the operation unit 32 of the report input device 30 by the interpretation doctor who is the user. However, the present invention is not limited to this. Various inputs may be realized by receiving voice to be emitted with a microphone and performing voice recognition. That is, various inputs may be performed in response to various user operations including operation of the operation unit 32 and pronunciation by the user.

  In the above embodiment, the information stored in the input support DB 111 or the structured DB 123 is created by analyzing an existing report by language processing. However, the present invention is not limited to this. Information stored in the DB 123 may be appropriately created manually. However, from the viewpoint of facilitating the provision of metadata, it is preferable that information stored in the input support DB 111 or the structured DB 123 is automatically input.

  In the above embodiment, the server 100 and the terminals 10 to 50 are connected so as to be able to send and receive data via a network line NTW in the hospital, and the interpretation report accumulated in the hospital is used as past knowledge. Although the support function is provided, the present invention is not limited to this. For example, as shown in FIG. 29, the server 100 has a service provider that supports the input of the interpretation report or a specific hospital, and the server 100 has an Internet line. The database system 1A may be configured such that an interpreting doctor at each hospital can access through a terminal (for example, the report input device 30A) and use an input support function including a search function.

  According to such a configuration, the user can obtain visible information on a similar image or related report for the instructed image using the Internet, which can easily increase the amount of information to be searched. Therefore, the probability of obtaining a desired image or report is further improved.

  At this time, a trader or the like having the server 100 may provide a service for storing images, interpretation reports, and the like of each hospital.

  However, when searching for similar images and related reports using the Internet, personal information included in attribute information attached to similar images and personal information described in the related reports may be leaked. A typical example of this personal information is a name (patient name or the like).

  Therefore, in the server 100, after a predetermined type of personal information is deleted from the detected similar image or related report data, information related to the similar image or related report is provided to the terminal 30. good. At this time, as a method of recognizing a predetermined type of personal information, for example, a method of recognizing character information belonging to a predetermined type of classification item “patient name” as a predetermined type of personal information can be considered. Character information related to the recognized personal information can be easily deleted.

  With such a configuration, personal information can be prevented from leaking over a wide range.

  In the above embodiment, various operations are performed in the server 100 in response to user input in the report input device 30. However, the present invention is not limited to this. For example, the functions of the report input device 30 and the server 100 are as follows. You may comprise by one computer apparatus. In addition, the report input device 30 is configured such that the control unit 120 of the server 100 substantially performs image data specification and search condition specification of the information search function only by a user performing various inputs. Also good.

  In the above embodiment, similar image and related report data are searched from the image data to be interpreted as the starting point. However, as a method for determining the starting image data, for example, the following modes are also considered. It is done.

  FIG. 30 is a diagram illustrating an input support template display TG1 according to a modification. In the input support template display TG1, in the same manner as the input support template display TP2 according to the above-described embodiment, the phrase groups FD1 to FD4 including one or more words are included for the items “imaging condition”, “part”, “basic findings”, and “diagnosis”. Each is displayed, and a search button SB is further added.

  The search button SB is a button for searching image data that is a starting point in data search for similar images and related reports. When the user performs various operations on the mouse to move the mouse pointer MP to the search button SB and press the left button of the mouse (left click), below the input support template display TG1 (that is, near the finding sentence display area A6) ) Displays a dialog box (search dialog) DG1 for determining search conditions as shown in FIG.

  In the search dialog DG1, by specifying words for each item “imaging condition”, “part”, “basic findings”, and “diagnosis”, each item “imaging condition”, “part”, “basic findings”, “diagnosis” and the respective items A combination of words belonging to the item can be specified as a search condition.

  For specifying the search condition in the search dialog DG1, in the state where the search dialog DG1 is displayed, the desired phrase included in the phrase groups FD1 to FD4 displayed in association with each item on the input support template display TG1. It can be executed by moving the mouse pointer MP and performing a selection operation (left click). In FIG. 31, for example, in the state where the search dialog DG1 is displayed, the words “T2FLAIR image” for the item “imaging condition” and the words “temporal lobe” and “frontal lobe” for the item “part” in the input support template display TG1. A state is shown in which search conditions are specified in which the phrase “high signal area” is associated with the item “basic findings” and the phrase “lacuna infarction” is associated with the item “diagnosis”.

  When the search condition is specified and then the mouse button MP is placed on the search button SB2 and a pressing operation (left click) is performed, the structured DB 213 is used as a search target and the attribute completely and / or partially matches the specified search condition. One or more single report structured data that is assigned a value is detected. Then, a dialog box (search result display dialog) DG2 for displaying a search result as shown in FIG. 32 is displayed in an area extending from the creation support information display area A5 to the finding sentence display area A6.

  The search result display dialog DG2 mainly includes an area for displaying a search result on the left (search result display area) AL and an area for displaying a search condition for performing a right search again (re-search condition display area). It is comprised by AR.

  In the search result display area AL, an image related to one or more single report structured data to which an attribute value that completely and / or partially matches the search condition specified in the search dialog DG1 (FIG. 31) is assigned. The data of the observation sentence is output visually. In FIG. 32, 146 single report structured data were detected, and images and findings data related to 5 single report structured data were output visually in order from the top. The state is shown. In the rightmost area of the search result display area AL, for each single report structured data, a score indicating the degree of coincidence with the search condition is shown. For example, the scores are detected in descending order of score. Single report structured data is visually output from the top in spatial order.

  By referring to the image displayed in the search result display area AL and the content of the finding sentence, the user can appropriately determine an image serving as a starting point for data search for similar images and / or related reports. Specifically, when the mouse pointer MP is set to a desired image among the images displayed in the search result display area AL and the right button of the mouse is pressed, the command as shown in FIG. List CL1 appears. For example, when a command “search for similar images” or the like in the list CL1 is designated, image data of a desired image combined with the mouse pointer MP among images displayed in the search result display area AL is obtained. The search result display screen as shown in FIG. 23 is displayed by the same processing as in the above embodiment, determined as the image data of the search reference image that is the starting point in the data search for similar images and related reports. In this similar image and related report data search, an image that has already been inspected is designated as a search reference image. Therefore, in the structured DB 213, the image data of the search reference image is directly and / or indirectly related. One or more pieces of character information specified are used as a search keyword group.

  In the search result display area AL, a number is assigned to the right side of each image, and this number indicates the name of the file storing the original image data in the image DB 212. When the mouse pointer MP is set to a desired number and a designation operation (left click) is performed, linking is performed so that the original image data is displayed visually large.

  On the other hand, in the re-search condition display area AR, the search condition for the re-search is determined in order to appropriately narrow down candidates for images that can be the starting point of data search for similar images and / or related reports and display them in the search result display area AL. be able to. As shown in FIG. 32, in the re-search condition display area AR, “patient name”, “patient age”, “examination date”, “part”, “modality”, “in charge” for the conditions specified in the search dialog DG1 (FIG. 31). Other conditions can be added for items such as “medicine” and “radiologist”. In addition, for each search keyword, a condition related to fluctuations of a phrase such as a compound word, a synonym, a synonym, etc. (for example, a compound word, a synonym, or a synonym to be considered for each word in a search) is appropriately added to the search condition You can also. Although not shown, when the mouse pointer MP is placed on the re-search button provided at the bottom of the re-search condition display area AR and a pressing operation (left click) is performed, the structured DB 213 is selected as a search target. , One or more single report structured data to which an attribute value that completely and / or partially matches the re-specified search condition (re-search condition) is detected, and the display contents of the search result display area AL Is updated.

  By adopting such a configuration, when the conditions specified in the search dialog DG1 (FIG. 31) are insufficient in narrowing down image candidates that can be the starting point of data search for similar images and / or related reports. The user can quickly and easily determine image data serving as a starting point for data retrieval of similar images and / or related reports by appropriately performing re-searching. That is, it becomes easier to input search conditions in data search for similar images and / or related reports.

1,1A database system 19 thesaurus 30,30A terminal (report input device)
31, 120 Control unit 32 Operation unit 33 Display unit 34, 110, 200 Storage unit 100 Server 111 Input support database 123 Data construction unit 125 Information search unit 130 Dictionary database 210 Medical information database 211 Report database 212 Image database 213 Structured database 312 Display control unit 313 Query generation unit 401 Image display screen 402 Input screen IN Internet line LN, NTW Network line

Claims (16)

A database system,
For each report included in a plurality of reports, related information in which a plurality of character information constituting the report and one or more image data corresponding to the report are directly and / or indirectly associated is accumulated. A related information database;
A plurality of character information items are listed for each attribute item included in a predetermined number of attribute items to which the plurality of character information items belong, and a plurality of character information items belonging to the respective classification items are between the predetermined number of attribute items. An input support information database in which network information associated with each other in a network form is stored;
Based on a part of the network information included in the network information, at least one character information is listed for each classification item for a plurality of classification items included in the predetermined number of classification items, and between the plurality of classification items. Output means for visually outputting a list display in which a plurality of character information belonging to each classification item is associated in a network form, and at least one image data;
Designating means for designating character information for each of the plurality of classification items in the list display in response to a user operation in a state where the at least one image data is visually output by the output means. When,
Determining means for determining character information designated for each of the plurality of classification items by the designation means as an element constituting a report according to a predetermined report model;
1 for one or more classification items which are in a state where the at least one image data is visibly output by the output means and are part of the plurality of classification items in the list display by the designation means. In the state where the above character information is designated, in response to the user's action, the one or more character information is used as a search keyword group, and direct and / or indirect for each image data in the related information database. By performing a keyword search using the plurality of character information associated with the metadata as metadata for data search, at least one or more image data related to the at least one image data visually output by the output means A report detection means for detecting report data;
A database system comprising: The database system according to claim 1, wherein
For each report included in the plurality of reports, a report database in which report information in which a report and one or more image data corresponding to the report are associated is accumulated;
Character information extracting means for extracting a plurality of character information constituting each report by performing language analysis on each report included in the report database;
For each report included in the plurality of reports, a plurality of character information constituting the report extracted by the character information extraction means and one or more image data corresponding to the report are directly and / or indirectly Storage control means for constructing a related information database by accumulating in association;
A database system further comprising: The database system according to claim 1 or 2, wherein
A plurality of character information constituting the related information,
A database system comprising attribute information given to one or more image data corresponding to each report. The database system according to claim 3, wherein
The attribute information is
A database system comprising at least one or more of information indicating patient ID, age, sex, modality, examination site, interpretation doctor, and doctor in charge. The database system according to any one of claims 1 to 4, wherein
The related information is
A database system characterized by being described in RDF. The database system according to any one of claims 1 to 5,
A dictionary database in which each character information included in at least one or more character information is stored in association with one or more related character information having a predetermined relationship with each character information;
Further comprising
The keyword search is
Including one or more related character information having the predetermined relationship with respect to one or more pieces of character information included in the search keyword group and adding the extracted information to the search keyword group. Database system. The database system according to claim 6, wherein
The predetermined relationship is
A database system comprising at least one relationship among synonyms, synonyms, and related terms. The database system according to claim 6 or 7, wherein
The predetermined relationship is
Including the hierarchical relationship of the concept of character information,
The keyword search is
A database system comprising: processing for extracting one or more superordinate character information from the dictionary database for one or more character information included in the search keyword group and adding the character information to the search keyword group. The database system according to any one of claims 1 to 8,
The predetermined number of classification items is
A database system comprising at least one item of imaging conditions, site, basic findings, characteristics, diagnosis, and conclusion. The database system according to any one of claims 1 to 9,
The keyword search is
A predetermined predetermined value included in the predetermined number of classification items among one or more pieces of character information specified for one or more classification items of the plurality of classification items in the list display by the designation means. The database system characterized by including the process which selectively employ | adopts the 1 or more character information which belongs to some classification | category items as said search keyword group. The database system according to any one of claims 1 to 10,
An item designating unit for designating at least a part of the predetermined number of classification items in response to a user action;
Further comprising
The keyword search is
Among the one or more pieces of character information specified for the one or more classification items of the plurality of classification items in the list display by the specification means, some of the classification items specified by the item specification means A database system including a process of selectively adopting one or more pieces of character information to which the search keyword group belongs. The database system according to any one of claims 1 to 11,
The plurality of reports are
A database system including an interpretation report. A database system according to any one of claims 1 to 12,
A server having the report detection means;
A terminal device connected to the server so as to be able to transmit and receive data via an Internet line;
With
The designation means is
In response to the user's operation on the terminal device, in the state where the at least one image data is visibly output by the output means, character information is respectively displayed for the plurality of classification items in the list display. Specify
The server is
Transmitting means for transmitting one or more report data detected by the report detecting means to the terminal device via the Internet line;
Have
The terminal device is
A database system comprising the output means, wherein the output means visually outputs one or more report data transmitted by the transmission means. The database system according to claim 13, wherein
The server is
Deleting means for deleting predetermined types of personal information from one or more report data detected by the report detecting means;
Further comprising
The transmission means is
One or more report data from which the predetermined type of personal information has been deleted by the deleting unit is transmitted to the terminal device via the Internet line.   A program for causing the database system to function as the database system according to any one of claims 1 to 14 by being executed by a computer included in the database system. A report search method,
(A) For each report included in a plurality of reports, related information in which a plurality of character information constituting the report and one or more image data corresponding to the report are directly and / or indirectly associated is accumulated. A plurality of pieces of character information are listed for each attribute item included in a predetermined number of attribute items to which the plurality of pieces of character information belong, and each classification item is included between the predetermined number of attribute items. Preparing an input support information database storing network information in which a plurality of character information belonging to each other is associated with each other in a network form;
(B) Based on a part of the network information included in the network information, at least one character information is listed for each classification item for the plurality of classification items included in the predetermined number of classification items, and the plurality of classifications A step of visually outputting a list display in which a plurality of character information items belonging to each classification item are associated in a network form between items and at least one image data by an output unit;
(C) In the state where the at least one image data is visibly output by the output means, in response to a user operation, character information is designated for each of the plurality of classification items in the list display. And steps to
(D) determining the character information designated for each of the plurality of classification items in the step (C) as an element constituting a report according to a predetermined report model;
(E) One or more classification items that are in a state in which the at least one image data is visually output by the output unit and are part of the plurality of classification items in the list display Directly and / or indirectly for each image data in the related information database using the one or more character information as a search keyword group in response to a user action in a state where one or more character information is specified. One or more related to the at least one image data visually output by the output means by performing a keyword search using the plurality of character information that are associated with each other as metadata for data search Detecting report data for
The report search method characterized by comprising.

Images